This PDF file contains the front matter associated with SPIE Proceedings Volume 12748, including the Title Page, Copyright information, Table of Contents, and Conference Committee information.

This paper introduces a weak optical signal processing scheme based on FPGA. Through the design of weak signal conditioning module, data acquisition control module, FPGA and peripheral modules, multi-channel acquisition, amplification, filtering and processing of weak light signals are realized. Finally, the feasibility of the scheme is verified through relevant experiments.

Tracking specific target is an important research area in machine vision. When tracking specific target encounters the same local features with light and dark variations, convolutional neural networks based on deep learning require large data sets to extract target feature information. To address these problems, this paper proposes a target of interest tracking scheme based on specific object imaging. The method allows for fast capture and processing of important information about the target of interest. In addition, small data sets are used to track the target and extract the target of interest efficiently. The target of interest can also be accurately tracked when the background and light changes. The results show that our method reduces the processing of redundant information, reduces the storage overhead, and tracks the target accurately.

In the Marine environment, underwater acoustic positioning signals are susceptible to complex underwater acoustic channels and multiple types of underwater noise, resulting in inaccurate DOA estimation, which will significantly reduce the accuracy and reliability of underwater positioning, and the real-time performance of underwater positioning cannot be guaranteed. In this paper, a Time-Frequency fusion based convolutional neural network (TFF-CNN) method for underwater DOA estimation is proposed. In this method, fusion feature maps are constructed based on Time-Frequency characteristics of signals, and a novel convolutional neural network with the deconvolution layer is designed to enhance feature representation and noise resistance. The experimental results show that compared with traditional methods and other networks, the proposed algorithm has great advantages in the estimation accuracy and speed, especially under non-Gaussian noise.

With the development of hardware resources, artificial intelligence has gradually emerged in the public domain, which can be well applied to the problem of image recognition. At present, a lot of exciting achievements have been made in the application of artificial intelligence algorithms to image recognition. This paper focuses on the research in the field of image recognition, improves the commonly used training algorithms of artificial intelligence, and proposes the research of image recognition system based on artificial intelligence for different application scenarios. Finally, the features trained in this study are applied to image recognition tasks to obtain better image retrieval results than those based on traditional image recognition methods.

The development of image processing technology has provided strong support for image recognition technology. At present, image recognition technology has gradually broken through the concept limit and has been applied to many fields. Image recognition technology is an innovation and upgrading of image processing technology, which is mainly used to collect and transmit various information through computer operation. At present, image processing technology mostly adopts the method based on depth learning model, but the traditional depth learning model has many disadvantages in image processing. In order to solve the problems of a single depth learning model, an image recognition system design based on depth learning hybrid model is propose.

The traditional research of power text information mostly uses manual input into the computer, and then uses machine learning or deep learning methods to further study the text. It can be seen that it needs to spend a lot of human resources to input the corresponding text information. In order to solve the above problems, the research of power text information system based on image detection and recognition under the MVC framework is proposed. First, the power text information is recognized using image detection technology, Convert to the form of digital matrix, and then extract the contextual semantic information in the digital matrix using the cyclic neural network. In addition, in order to further improve the effect of information extraction, the attention mechanism is introduced in semantic information extraction, which focuses on the words that have a great impact on the final result, so as to improve the effect, and then the MVC architecture is used to design and implement the final information recognition system, The experimental results show that the proposed power text information system based on image detection and recognition under the MVC framework can effectively improve the effect of text information research.

Motion modeling and temporal modeling are crucial issues for video behavior recognition. When extracting motion information in two-stream network, the optical flow diagram needs to be calculated in advance and the end-to-end training cannot be realized. 3D CNNs can extract spatiotemporal information, but it requires huge computational resources. To solve these problems, we propose a plug-and-play motion capture and enhancement network (MCE) in this paper, which consists of a temporal motion capture module (TMC) and a multi-scale spatiotemporal enhancement module (MSTE). The TMC module calculates the temporal difference of the feature-level and captures the key motion information in the short temporal range. The MSTE module simulates long-range temporal information by equivalent enlarging the temporal sensitive field through multi-scale hierarchical sub-convolution architecture, and then further enhances the significant motion features by referring to the maxpooling branch. Finally, several experiments are carried out on the behavior recognition standard datasets of Something-Something-V1 and Jester, and the recognition accuracy rates are 49.6% and 96.9%, respectively. Experimental results show that the proposed method is effective and efficient.

In distributed passive localization and tracking system, the track observed by the subsystem seems like Brownian motion track, because the tracked target is non-cooperative target and its maneuver is often complex, and the localization accuracy is poor. These track characteristics will seriously disturb track association between different subsystems. In order to solve this problem, the track to track association algorithm based on empirical mode decomposition (EMD) is proposed in this article. To lessen the impact of target placement and maneuvering mistakes, components that do not follow the track trend are removed from each dimension of the track recorded by each sub-system. The track motion trend vector is formed using the remaining low-frequency components as track characteristics, and the relevant correlation criteria are created. The track association between sub-systems is ultimately finished since the correlation threshold is self-adaptive and does not require the creation of a motion model. Results from simulations indicate that the suggested method is capable of successfully completing the track connection in passive systems

In radio spectrum technology, the modulated signals are susceptible to multipath interference, the bit error rate is increased under multipath conditions, and the signal is distorted. The article proposes a modulation signal transmission interference suppression method based on passive time reversal mirror technology in radio spectrum technology, the interference signal is suppressed during transmission to realize lossless transmission. Firstly, a multipath transmission channel model is established in information transmission. Fractional interval balanced technology is used for channel equalization design, and passive time reversal mirror is used for intercode interference suppression and blind signal separation, The lossless transmission of modulated signals is improved by passive time reversal mirror method in wireless spread spectrum communication system. Finally, System performance is tested by simulation method. The channel balance performance is better, the intercode interference is better, and communication symbol error is lower than conventional method.

To improve the performance of high-frequency information acquisition data processing and meet the differentiated power service requirements in low-voltage distribution grid, this paper first investigates an edge processing architecture for high-frequency information acquisition in low-voltage distribution grid. Then, an edge processing mechanism for high-frequency information acquisition is proposed to optimize the high-frequency information acquisition data distribution strategy and communication and computation resource allocation strategies. Simulation results demonstrate that the proposed mechanism achieves better data processing performances compared with traditional data processing mechanism

In order to realize high-frequency information acquisition, differentiated information transmission, and real-time processing for low-voltage transparent substation area, this paper first proposes an architecture of edge-device collaborative high-frequency information acquisition. Then, a data priority-aware matching-based high-frequency information acquisition data transmission strategy optimization method is proposed to improve weighted data offloaded amount by adopting data priority-aware matching. The simulation results show the effectiveness of the proposed algorithm.

Ultraviolet and infrared images are widely used to detect potential defects of electric equipment in advance. However, With the continuous expansion of the scale of electrical equipment, almost all the data is still uses manual analysis, which has some shortcoming example high cost. In order to improve the intelligence level of electric equipment detection, this paper proposes a multiple images identification method for electrical insulator based on two stage cnn network. First, a deep convolutional neural network identification model is built based on Faster-RCNN with initial weights in VGG. In addition, combine with the dataset expansion method example rotation, 462 images of electrical insulator are used for training. Finally, this model is used to identification image of equipment. The experimental show that the method proposed in this paper for electrical insulator map accuracy reaches 96.51%. Therefore, the proposed method can improve the identification accuracy, and provides condition for electrical equipment state detect.

Photon counting detector (PCD) is a hot topic at present. Compared with traditional energy integral detector, it has the potential of high spatial resolution, high sensitivity and low dose, which can effectively promote medical imaging diagnosis. However, when PCD is counting X-ray photons, the photon number of each energy bin is relatively small. Additionally, charge-sharing response and pulse superposition effect will also affect the photon count rate, resulting in serious noise and affecting the imaging quality. In this paper, a photon-counting denoising algorithm based on subspace decomposition is proposed. According to the similarity between the data of different bins and the self-similarity of the data, this paper constructs sparse representation by subspace decomposition method and uses block matching algorithm to suppress noise. In simulation experiments, we carried out spectral computed tomography imaging experiments with the three-dimensional phantom of a digital mice based on PCD, and denoised the data by different algorithms. The quantitative results show that our method improves peak signal-to-noise ratio by 2.21dB compared with block-matching and 3D filtering when photon flux is 4×10³ , which verifies the potential of the proposed algorithm in medical imaging.

X-ray tomographic imaging has become an important analytical tool with a wide range of applications. It is inevitable that noise is introduced in CT images, and noise reduction is necessary. To solve this problem, we considered to use the nonlocal property of similar block search and proposed a deep learning network based on similar block learning for noise reduction of micro CT short exposure time scanned images to improve the scanning efficiency while ensuring high quality imaging. The method uses the output of the nonlocal method as a data preprocessing algorithm by combining a nonlocal block matching algorithm with a convolutional neural network, and uses a residual channel attention mechanism to learn the features after feature extraction, which reduces noise while preserving image details. Experimental results show that the method can remove noise from CT images quickly and effectively, and compared with the classical CPCE noise reduction method, the method improves the PSNR index by 1.52 dB, which is consistent with the theoretical assumption.

In the explosion experiment, the explosion vibration signal is the important information to evaluate the explosion equivalent or to locate the explosion point. However, the environment at the scene of the explosion is very complicated. The explosion vibration signal collected by the Distributed fiber Acoustic Sensor (DAS) system not only contains optical noise, but also contains a lot of non-stationary and non-Gaussian environmental background noise. In order to solve this issue, feed-forward denoising convolutional neural network (DnCNN) are used to denoise the signals during the noise suppression of explosion vibration signals. The initial application of this network was to purge images of additive Gaussian white noise. In order to make DnCNN adapt to the de-noising of explosion vibration data, we have performed numerous optimization tasks. Firstly, the input data is processed by an reversible downsampling factor in this paper, expanding the network’s perceptual field while also making training easier. Secondly, rather than using Gaussian white noise, the training set of DnCNN is rebuilt using background noise collected in the actual environment. Finally, the DnCNN parameters that impact network performance are modified and improved. From the experimental results in this paper, the DnCNN can effectively suppress the noise in the explosion vibration signal and preserve the effective signal.

With the development of artificial intelligence, the fault detection of substations has changed from manual to artificial intelligence detection. The higher the image quality, the better the recognition accuracy of the model, but when the inspection image is acquired, transmitted and saved, it is usually affected by bad weather, relative motion, imaging equipment shaking and other factors, which makes the acquired image blurry. Therefore, a fuzzy image screening model is constructed to screen and remove blurred images. In this paper, the grayscale co-occurrence matrix is used to extract the characteristics of the texture feature information of the image, extract the four feature values of the image, take the feature values as the input of the MLP neural network, and use the TID2013 dataset for the training dataset, and finally realize the quality scoring and screening of the blurred image.

The integrated construction of Bluetooth positioning network and 5G digital indoor division system can share power, transmission and other resources, which is convenient for deployment and maintenance, and provides high-accuracy indoor positioning capability for various types of terminals such as smart phones and Bluetooth tags. In this paper, we analyzing the key issues of 5G/Bluetooth AoA hybrid positioning, and propose a hybrid positioning solution of 5G/Bluetooth AoA/terminal motion sensor, effectively improves the positioning accuracy and applicable scenario through complementarity of positioning technology.

Aiming at the problems of low recognition rate of complex shape defects on wind turbine bearing surface and inaccurate localization of small target defects by traditional inspection methods, this paper proposes an improved SOLOv2-based bearing surface defect recognition method. The method selects ResNeXt-101-FPN as the backbone network on the basis of SOLOv2 model framework, combines deformable convolution (DCNv2) to strengthen the learning of defect morphological features by the network, which is more adaptable to complex defect objects; proposes improved mask feature branching, adds adaptive attention module and feature enhancement module, which not only reduces the loss of feature information, but also enhances the feature expression and change the way of feature fusion. Experiments show that the mean average precision (mAP) of the improved SOLOv2 network for bearing surface defect recognition is 96.6%, which is 2.8% higher than that before the improvement. The average recall (AR) is 96.8%, which is 1.6% higher than before the improvement; the improved model is less affected by the variable shape of the defects, locates small targets more accurately, and effectively solves the problems of partial missed segmentation and under-segmentation.

Image feature point extraction and matching are important branches in the field of computer vision. Based on the extracted image feature points, the image descriptor is calculated, and the matching information between two images is generated using the image descriptor. The relative position relationship between the two images is then recovered using homography transformation, which is the basis of monocular visual odometry. In this study, the Key.Net method is used to extract image feature points, and SIFT and SURF methods are used to calculate feature point descriptors. The images are matched using a brute-force matching method, and the relative position relationship between the two frames is solved. Multiple image feature point extraction and descriptor calculation methods are compared on the KITTI dataset using the distance information between two frames as the error benchmark. The proposed method has the lowest total error and the best performance, and it has good robustness, laying a foundation for visual navigation algorithms.

Aiming at the problem that the estimation accuracy of the direct position determination (DPD) method based on subspace is not high under the condition of small number of snapshots and low signal-to-noise ratio (SNR), a noncircular signal DPD method based on off-grid sparse reconstruction is proposed. This method combines the non-circular characteristics of the signal to expand the received data and then expand the array aperture. Then, based on the spatial sparsity of the target location, an ultra-complete dictionary set is constructed by discretizing the location area grid, and the problem of target position estimation is transformed into the problem of spatial signal sparse reconstruction. At the same time, considering the signal model that the target is not on the grid point, the joint optimization problem is solved by the alternating iteration method to obtain the estimated value of the target position. Finally, the experimental simulation shows that the method has better positioning performance.

Image deblurring refers to the process of restoring sharp images from blurry images, which is one of the pivotal technologies in many domains. In this paper, we introduce a novel neural network, which contains DIP network, double blur kernel processing network and blur kernel size estimation network. First, DIP network mainly extracts the prior information of blur kernel and sharp image. Then, double blur kernel processing network conduct similarity processing on a pair of blur kernel information collected. Compared with state-of-the-art methods. our method has a great improvement in running time and deblurring performance

To process broadband signals and analyze their power spectrum, it is necessary to consider the parallelization, accuracy and system complexity of processing. In this paper, a wideband signal power spectrum analysis method based on Fast Fourier Transform (FFT) is proposed and implemented on the platform of Field Programmable Gate Array (FPGA). This method realizes the FFT of high-speed serial signals by parallel computing, so it can realize the power spectrum analysis function of signals when the sampling rate is much higher than the processing rate. At the same time, according to the processing characteristics of FPGA IPCore, two real FFT operations are realized by one complex FFT operation, so as to minimize the resource utilization. This paper completes the simulation analysis on Matlab, and implements it on the hardware platform of EV12AQ600 ADC and Xilinx xc7vx690tffg1927-2, and finally presents the consistent results of the two.

Cross-eye jamming is a kind of coherent jamming, which can effectively interfere with monopulse radar in angle measurement. On the basis of discussing the principle of cross-eye jamming technology, we deduce the calculation steps of using three jammers to generate spatial phantom targets, and simulate the jamming technology in combination with a practical application scenario. The simulation result shows that this jamming technology can generate phantom tracks in three-dimensional space and effectively interfere with monopulse radar.

A lightweight network based on YOLOv5 is proposed in this paper to improve the real-time detection ability of underwater targets for fishing gear and to solve the difficulty of deploying model algorithms on embedded devices. First, the Shuffle_Block module replaces the leading feature extraction network in YOLOv5, reducing parameters and improving the algorithm's inference speed. Second, this module is combined with depthwise separable convolution to construct the feature fusion Shuffle-PANet, significantly reducing network parameters and improving detection speed while ensuring accuracy. The proposed method in this paper has been verified to reduce the parameter count by 89% compared to the YOLOv5 source code while doubling the detection speed of the source code. Additionally, the weight file size is reduced by 83%. The mAP50 reaches 96.3%, which is only a 2% decrease compared to YOLOv5. The lightweight network proposed in this paper can recognize sea cucumbers well and has fast recognition speed and lightweight design characteristics. Shuffle-YOLOv5 has significant advantages compared to the original model and can complete real-time target detection on low-power embedded devices.

With the widespread use of recommendation systems, researchers have gradually started to focus on fairness issues including two-sided fairness along with recommendation accuracy. However, these works often consider the relationship between the provider and the recommended items as one-to-one or one-to-many, without considering the situation that each item may have multiple related individuals on the provider side. In this paper, we implement a two-sided fair perception recommendation method based on fair allocation in a context where the relationship pattern between providers and items is many-to-many relationship. Specifically, on the one hand, the attention of all consumers is considered as the total exposure available to the provider and is fairly allocated to each provider based on the fairness criterion, and on the other hand, each exposure of each item is allocated to each relevant provider based on the fairness criterion, and the sum of the exposure obtained by each provider in all relevant items is taken as its final exposure. We conducted experiments on a real-world dataset, and the results show that the approach in this paper provides better two-sided fairness compared to the benchmark approach while maintaining good recommendation quality.

In recent years, information security has become more and more important. Image encryption technology based on chaos theory has become one of the hot issues in this field. Chaotic system has the characteristics of initial value sensitivity and sequence ergodicity, which is very suitable for image encryption. In this paper, the folded surface is generated by connecting the randomly generated points on the bottom surface and the top surface in the unit space, and the surjective binary function is further constructed. Use this type of function and functions such as planes to construct discrete dynamical system. It is experimentally analyzed that the function has good chaotic characteristics by drawing bifurcation diagram and Lyapunov exponent diagram. The chaotic sequence of the discrete dynamic system is used for image encryption, and its information entropy and correlation coefficient before and after encryption are calculated. It is proved that this kind of system has good chaotic characteristics. This is a new type of chaotic system, which needs further research, analysis and expansion.

In this paper, we propose a Swin Transformer-based Asymmetrical Network (SwinAN) for light field salient object detection (SOD). SwinAN is a bifurcated backbone network composed of a 2D branch and a 3D branch. The 2D branch is driven by Swin Transformer blocks for all-in-focus images, while the 3D branch is constructed by a 3D convolutional neural network (CNN) for focus stacks. Then, the high-level features extracted by the asymmetric backbone network are integrated using cascaded Multi-Level feature Fusion (MLF) modules, which consist of channel attention mechanisms and a residual block. Finally, the Multi-Attention Fusion (MAF) module is used to fuse multi-modal features to generate the predicted saliency maps. Experimental results show that after comparing 10 different SOD models, our SwinAN achieves excellent performance on 3 datasets, confirming the superiority as an efficient and accurate method for detecting saliency

With the rapid development of information technology, biometrics has been widely used in various fields of life.Among them, finger vein recognition technology, as a new technology in the field of biometrics, has been widely concerned because of its advantages of good security, high recognition accuracy and low requirements for acquisition conditions.In view of the low accuracy of the existing finger vein recognition and the influence of many factors such as finger rotation, translation and inaccurate placement, this paper optimizes the finger vein authentication technology by improving the algorithm: using three-dimensional vein authentication research, the finger vein pattern is obtained from four directions, which effectively overcomes the problem of incomplete pattern extraction.The contour of finger vein is extended to make better use of and transfer the shallow texture features and improve the recognition speed.The multi-feature fusion matching algorithm is used to greatly improve the robustness of the a lgorithm.According to the comparison of experimental data, the finger vein recognition technology algorithm in this paper has certain advantages and good application prospect.

In the applications of wireless sensor networks (WSNs), it is important to locate an object of interest. However, the intensive measurement noise that contaminates the observations from each sensor node, may impair the localization performance. The widely studied adaptive and cooperative schemes combat the noise via reliable cooperation and adaption strategies with the neighborhoods. However, they underestimate the smooth correlations of the object’s movements, thereby remaining space for improvement. In this paper, we focus on improving the existing cooperative schemes by prefiltering its contaminated observations on each node. By exploiting the smooth correlations of the object’s mobility, we design a sequential pre-filter, which is capable of using the previously estimated information as a priori to overcome the intensive noise. As such, it helps to derive a less-noisy observation on each node, and therefore improves the localization accuracy of the cooperative schemes. Numerical simulations demonstrate the effect of the proposed sequential pre-filter, which can indeed better the cooperative schemes and gain a more promising localization performance.

Ground Penetrating Radar (GPR) data requires a significant amount of network bandwidth and storage space for transmission and storage due to the large number of channels and vast amount of data. In this paper, we propose an improved method for compressing GPR data. Firstly, we analyze and preprocess the features of the data to enhance its compression potential. Secondly, we introduce convolutional layers into the AutoEncoder to improve its generalization ability. We then use multiple-level compression to further compress the data based on the radar data's features. Finally, we introduce range encoding for secondary compression. Simulation experiments demonstrate that our proposed algorithm can effectively compress radar data while maintaining high compression ratios and speed.

With the development of economy and society, people's requirements for power quality and reliability are gradually increasing, so it is necessary to carry out risk assessment on power system. Monte Carlo method is a common method for power system risk assessment, but there is a contradiction between the calculation speed and accuracy. In order to improve the calculation accuracy, a large number of state samples need to be calculated, which takes a long time, and there are many repetitive states. For this reason, an improved Monte Carlo method based on Huffman code and state identification is proposed, which can significantly improve the efficiency of power grid risk assessment from two aspects. The first is to uniquely identify the system states through Huffman code and record relevant data, so as to avoid recalculation of the same states. The second is the fast and efficient identification of system state based on the shortest weighted path length of Huffman code. The effectiveness of the method is verified by an example, and several factors that may affect the effectiveness of the method are analyzed.

Speaker recognition, also known as voiceprint recognition, is a biometric technology with wide practicability at present. This paper summarizes and compares and analyzes the main research methods of speaker recognition at home and abroad at this stage, and proposes an improved ECAPA_TDNN algorithm. It is proved by experiments that the improved ECAPA_TDNN algorithm in this paper is superior to the classical algorithm in terms of accuracy and loss.

Environmental sensing is an essential aspect of autonomous driving systems, with millimeter wave radar currently gaining attention in academic circles due to its unique physical properties that complement optical sensing techniques such as vision. Compared to cameras and LIDAR, millimeter wave radar is not limited by light and meteorological conditions, boasts high penetration capabilities, and can operate around the clock to identify objects. However, the larger wavelengths of millimeter wave signals present significant challenges such as sparse point clouds and multipath effects, resulting in lower accuracy in environmental sensing. To address this issue, this paper proposes a point cloud enhancement method based on a GAN-LSTM network that converts the sparse point cloud data into semantically informative RF images, thereby improving object recognition accuracy. The proposed method is evaluated on the CARRADA dataset, and the experimental results demonstrate an improvement in object classification accuracy by 7.86% compared to the current state-of-the-art methods. This approach can significantly enhance the accuracy of millimeter wave radar-based environmental sensing in autonomous driving systems, enabling safer and more reliable vehicle operation.

High altitude areas often lead to more serious corona discharge, and produce more serious electromagnetic environment problems such as audible noise and radio interference, which greatly restrict the selection of transmission lines and cause certain environmental problems. In order to study the characteristics of audible noise in high altitude areas, Yangbajing Town, Lhasa, Tibet Autonomous Region, China (altitude :4320m) and Wuhan city, Hubei, China(altitude :20m). The audible noise characteristics of three different lengths of discharge fine copper wires in high altitude areas (0.8mm, 1.0mm, 1.2mm) were tested. The influence of applied voltage and wire diameter on corona audible noise in plain area and high altitude area is analyzed. The research can provide important data reference for the study of electromagnetic environment in high altitude areas

The decision-making trial and evaluation of laboratory method (DEMATEL) can effectively analyze the relationship between the intrinsic factors of complex systems, so it is widely used in many fields. In this paper, based on a new ordered weighted average operator of three-parameter interval grey number (TPIGN-OWA), a novel DEMATEL model is given, which has three advantages: The first one is that the three-parameter interval grey number (TPIGN) can reflect the real intention of the decision maker more than the interval grey number. The second one is that the new TPIGN-OWA takes into account the weight of the two intervals assembly, thus, the intention of experts can be more accurately expressed. The last one is that in the decision-making process, it is not necessary for the decision-maker to have strong mathematical knowledge background, but only to give the corresponding TPIGN according to the scale of the importance of factors, which is easier for the decision-maker to operate. Finally, this method is applied to the evaluation of fire safety management, and its the effectiveness and practicability is verified.

In multi-attribute decision making, the aggregation method of each attribute value is always one of the core issues of research. This paper presents a novel aggregation operator as the further study to ICOWA operator. Firstly, the paper gives the method of converting interval number into binary connection number, and achieves the ranking of interval number by calculating the set pair potential of binary connection number. On this basis, a new induced interval number weighted averaging operator based on set pair potential (SP-ICOWA) is given, and it is proved that this operator has good properties such as boundedness, idempotence, commutativity. Finally, an example of the evaluation of the fire fighting capability of a ship is given. The continuous ordered weighted averaging operator (COWA) and SP-ICOWA are both used to calculate the fire fighting capability of a ship, the result manifests the feasibility and validity of SP-ICOWA operator.

Since the existing research focusing on the measurement of periphery is basically on the definite periphery, the research on the uncertain periphery is rarely involved. The following work has been carried out in this paper. Firstly, the concept and algorithm of interval-valued vague set (IVVS) of periphery element’s guard degree and exchange degree are given; Secondly, using the interval-valued aggregation operator, the calculation method of the total guard degree and the total exchange degree of the periphery elements based on the interval-valued vague set is given. By calculating the security degree and the survival degree of the periphery set, the quantitative evaluation value of the periphery is obtained, and the model of using the periphery to evaluate the system is given. Finally, an evaluation example is given to verify the effectiveness and practicability of the evaluation model.

External pipe weld inspection is a crucial responsibility in today's industrial manufacturing. Traditionally, the image to be detected is generated by a camera that takes images from different angles of the outer pipe and then analyses the detection frame by frame, generating a large amount of data and leading to duplicate identification. In this paper, we propose an algorithm for stitching together a panorama of external pipe welds by first acquiring a sequence of images taken 360° around the same scene, then correcting the image of the pipe column surface by the column surface inverse projection algorithm, then matching the feature points by the Flann algorithm to complete the alignment of the two images, and finally seamlessly stitching and fusing the aligned images to obtain a panorama of external pipe welds. The experimental results show that the algorithm can effectively, quickly and accurately generate a panoramic stitched image of the external pipe weld column surface, which can further provide input for weld quality analysis.

This study propose an integrated user behavior and value profiling method, which aims to build a more comprehensive index system to evaluate user value and demand. This method mainly includes a Time-Decay-Based Label Matching (TDBLM) method and an extended RFM model. The TDBLM method can effectively deal with the multi-label problem of user charging records due to clustering, and can derive a unique label that can better characterize the current state of users; the extended RFM model is used for user value profiles, and the experimental results show that user value profiles are defined into four categories: Lost, Potential, Retention, and High-value. In order to evaluate the performance of K-means++ used in this method, its performance was compared with that of GMM and Mean Shift clustering methods, and the results showed that the clustering results of K-means++ have better stability and are more suitable for user profiling. Finally, this study also proposes a method to visualize user profiles and make the profile results as knowledge graphs to realize the visualization and fast retrieval of profiles and expand the profile function.

The bat algorithm (BA) was a kind of meta-heuristic algorithm that was simple and efficient in optimization, it had been widely applied in various fields. To improve the capability of the original BA, two methods used frequently in improvements were proposed in this paper: the gradient and sub-gradient methods, together with the Levy flights. Simulation experiments were carried out on the representatives of unimodal and multimodal benchmark functions. Results confirmed that not all traditional improvements were effective, some of the improved BA even work worse than the original one. However, the Levy flights could be considered a better replacement of randomness in applications.

Power prediction of photovoltaic generation is very important to ensure the safety of the power system. It can assist the scheduling of smart grid and the operation and maintenance of photovoltaic stations. In order to improve the accuracy of photovoltaic power prediction, a temporal convolution network (TCN) model is constructed, which is combined with adaptive white noise complete integrated empirical mode decomposition (CEEMDAN) and wavelet threshold denoising. In the data preprocessing, the sequence of sample data is decomposed as modal components of different time and frequency scales with CEEMDAN method and the high-frequency modal components are denoised through wavelet threshold. By reconstructing the denoised components, the noise-reduced data set is obtained. In the process of TCN construction, the sample data used for modeling is from DKASC photovoltaic power generation public dataset. The experiment of modeling results show that the designed model can improve the accuracy of power prediction compared with the traditional prediction model. And the modeling method has feasibility and effectiveness.

ViBe (Visual background extractor) algorithm is a motion target detection algorithm by background modeling, uses domain pixels for background modeling and updates the background model in real time. However, the traditional ViBe algorithm will produce disadvantages such as ghosting and shadow part false detection. This paper solves the problem of ghosting by image fusion mechanism, through the characteristics of the shadow itself, using the shadow part of the three-channel variance value for shadow detection and eliminate the shadow. The experiments show that our algorithm can quickly and effectively solve the ghost and shadow problems in the ViBe algorithm

The wide variety of 5G multi-scenario hybrid power data leads to the problems of poor encryption security and high time cost in traditional data encryption methods. To ensure the 5G network security of power data, a hybrid 5G classified and hierarchical electric power data encryption method in cloud storage with multiple features is proposed. By collecting multi-source power data information in cloud storage space and pre-processing it, multiple features in the data can be extracted and fused, and different types of data will be encrypted. SM4 is used to encrypt the data transmitted by the acquisition system, and the corresponding ciphertext is generated. Based on the feature fusion results of power data, the classification and grading of power data are realized. Simulation results show that encrypting and protecting power data through the proposed method can effectively reduce memory consumption and has better practicability in the actual situation of resource efficiency.

Aiming at the problem of insufficient accuracy of existing models in the process of air quality index prediction, an air quality index (AQI) prediction model (CALSTM) incorporating a multi-head self-attention mechanism is proposed based on Long Short-Term Memory network (LSTM). The model effectively extracts low-dimensional features of air pollutant concentrations and meteorological data related to the air quality index through Convolutional Neural Network (CNN) and uses LSTM to fully reflect the long-term historical process in the input time series. To improve the acquisition ability of the global information of the model context, the multi-head self-attention mechanism is used to extract the hidden information of the time series at different levels and improve the model’s generalization ability. In addition, to further improve the prediction accuracy of the model, an AQI time series difference method is proposed based on the data correlation analysis. The experimental results show that the CALSTM using the AQI time series difference method has achieved an effect of 16.27% on MAPE, and on the indicators of MAE, MSE, RMSE, and R2, they are 6.99, 114.79, 10.71, and 0.9586, respectively. Compared with LSTM has achieved better prediction results.

As the complexity of electricity consumption increases, the traditional distribution network is increasingly unable to cope with the complex distribution network load. In order to improve the economic efficiency of the distribution network, this paper proposes an active distribution network economic optimisation dispatching method based on an improved particle swarm algorithm for accessing virtual power plants. After combining the actual situation of a region, the combination of the types as well as the number of distributed power sources and energy storage systems within the virtual power plant is comprehensively considered. The IEEE33 node system is introduced for simulation analysis, and constraints are set and modelled according to active power, reactive power and load demand. By improving the particle swarm algorithm, the selection of inertia weight is optimised to control the scheduling of internal and external power output of the virtual power plant for the twenty-four hours of the day. At the same time, power is purchased from the upper grid in combination with the tariff, and finally the minimum daily operating cost of the distribution network is obtained. This method reduces costs by 11.4% compared to the non-optimised period. It also improves the speed of convergence and perfects the composition of the active distribution network.

With the continuous improvement of the switching frequency of power electronic devices, the stray inductance parameters in the converter circuit and the di/dt caused by the rapidly changing current cause large voltage spikes at both ends of the switch tube. Once the rated value is exceeded, the irreversible breakdown of the IGBT will be caused. At the same time, the high-frequency stray parameters also introduce large high-frequency noise, which is detrimental to the output performance. Therefore, it is necessary to extract the line stray parameters through the analysis of the IGBT switching process. The traditional extraction method regards the current change rate near the moment of the maximum peak voltage Ucemax as a fixed value, and calculates the stray inductance value using the voltage-current relationship, but its extraction accuracy is not high. This paper focuses on the overall stray parameters of the inverter system and considers that the di/dt is constantly changing during the switching process. The switching time period applicable to the extraction of stray parameters is analyzed and obtained. The stray inductance value of the converter circuit is obtained by the method of parameter fitting, and the accuracy of the parameter extraction method is verified by simulation.

In order to improve the accuracy of PM_2.5 concentration prediction, a CNN-GRU deep learning model based on fusion of Luong Attention is proposed. Firstly, the correlation between various air pollutants and meteorological factors and PM_2.5 concentration is comprehensively analyzed, and the high correlation data is formed into a feature set. Secondly, the feature set is input into CNN for feature dimensioning, and then the output results of each time step are extracted through GRU. Finally, by introducing the Luong attention mechanism, the attention scores of the hidden states at each position of the output sequence are calculated, and the context vector is weighted to highlight the input step information that plays a key role in the prediction of PM_2.5 concentration. The results show that using the CNN-GRU model with attention mechanism to predict the PM_2.5 concentration in the next 24 hours, compared with the machine model and other deep learning models, RMSE and MAE have a certain decline, and have a higher generalization ability.

In the heterogeneous ultra-dense 5G network environment, a more flexible network selection algorithm is needed to improve user quality of service (QoS) due to the intensive network deployment and the diversity of user terminal business types. In this paper, a network selection algorithm based on fuzzy-gray ideal solution is proposed, and the decision parameters are dynamically selected according to service QoS requirements. At the same time, without fuzzy reasoning, a gray ideal solution network scoring method based on threshold judgment is designed by combining gray correlation with ideal sorting algorithm. Simulation results show that the proposed algorithm can effectively reduce switching delay and improve the accuracy of network selection.

In the wind turbine drive chain test platform using hydraulic loading system to simulate the load on the wind turbine, so as to complete the fatigue test test of the key components of the wind turbine on the ground. The traditional method of using microcontroller or PLC as the core of the motion controller has been unable to meet the needs of modern society's high-precision industries, the new open motion controller based on PCI bus control has gradually developed into the mainstream of the market choice, PC and motion controller control mode is widely used in various industries of manufacturing. In this paper, LabVIEW software is selected as the control system development platform to study the implementation of the control method of Advantech's PCI motion control card, while the collected data is displayed and processed by a data acquisition card. According to the functional requirements, the human-machine interaction interface of the software is designed for data acquisition and storage. According to the interaction characteristics between modules, the detailed design and implementation of the back panel was carried out for different modules using different design modes.

In order to enhance the ability of the robot operating system to execute multiple tasks in parallel and fully exploit the parallel performance of multi-core processors, a method to optimize the multitask scheduling strategy on multiple cores is proposed, namely, the elimination-based fireworks algorithm based on multidimensional Manhattan distance. An edge sparkle is introduced in EFWAMMD to enable each iteration of the fireworks algorithm to select the fireworks at the most edge position in the feasible domain in order to improve the optimal value search domain range, and an adaptation value calculation method based on the multidimensional Manhattan distance is proposed as the multicore task scheduling adaptation value to narrow the search range of the fireworks algorithm and use the elimination roulette rule to achieve efficient task scheduling and optimise the problem of high complexity and long running time of traditional fireworks algorithms. Comparison experiments with GA and FWA algorithms show that EFWAMMD has a significant improvement in the efficiency of task scheduling and can improve the task processing throughput of multicore processors

In this paper, from the four main steps of topology transformation, network topology analysis, equipment modeling and data generation strategy, the CIM/XML data including AC/DC system is converted to the input data of power flow calculation. Firstly, from the perspective of switching topology between devices, Depth First Search algorithm (DFS) search and device topology splicing are carried out to realize the conversion from switch/node model to bus/branch model. Secondly, after the active topological islands are screened and the non-live devices are eliminated, the influence of the converter modeling in CIM/XML on the selection of AC and DC nodes is emphatically analyzed, and then the selection rules of DC nodes and the strategy of DC data generation with universality are proposed. Finally, taking the CIM/XML data derived from the Southern power grid of China with a voltage of 500kV and above as an example, the power flow calculation results are compared with the measured SCADA data to verify the effectiveness of the proposed strategy.

With the rapid development of big data and artificial intelligence, machine learning algorithms have become a popular topic in fault diagnosis and preventive maintenance. This paper focuses on predicting vehicle maintenance and proposes a constrained-time-based model using a Seq2Seq neural network structure based on GRU (gated recurrent unit) to extract the dependency relationship between vehicle maintenance time series data. Our proposed method effectively addresses the problem of the traditional method ignoring the temporal nature of vehicle maintenance data and enhancing the model's generalization ability. We also introduce an attention mechanism to automatically obtain key input time points significantly related to the current prediction output. Furthermore, we expand the model's input by concatenating other features of the vehicle attributes as constraints with the decoder's output to help the model better understand the input sequence. We conducted comparative experiments using real-world data. The results demonstrate that the constrained Seq2Seq model has better prediction performance than the traditional method, indicating that our proposed method has promising application prospects in vehicle maintenance prediction

Accurate prediction of fire environment changes is helpful to accurately grasp the development trend of fire and ensure the safety of personnel. It is difficult to establish an accurate prediction model because of the coexistence of multiple parameters, complex coupling relationship, time series and nonlinearity of fire environment. In this paper, long shortterm memory network model (LSTM) based on improved Harris Hawk algorithm (CHHO) is proposed to achieve accurate prediction of fire environment data. Then, CHHO is used to optimize the hyperparameters in LSTM, and the fire temperature is predicted based on the optimized parameters. The experimental results show that the method of CHHO automatic parameter selection solves the problem of manual selection of LSTM model parameters and gives full play to the best performance of the model. The five environmental parameters of indoor fire temperature was predicted. The average fitting effect of CHHO-LSTM reached 94 %. The results show that the model has high prediction accuracy.

Muons produced by cosmic rays can be used to reconstruct images by analyzing their energy and angle information after passing through a medium. These subatomic particles have strong penetrating ability and are sensitive to high-Z (high atomic number) materials, making them ideal for large-scale structural imaging and nuclear material detection, which is critical for maintaining nuclear safety. However, muon tomography faces challenges such as low natural muon flux and difficulties in image reconstruction. Therefore, developing effective imaging reconstruction algorithms is crucial for muon tomography. In this study, we present modifications to the ASR algorithm, then apply the modified version to experimental data. Our results show that the images reconstructed using the modified ASR algorithm exhibit good quality, indicating the algorithm's effectiveness.

Ground based telemetry stations are usually used to acquire real-time information of flight vehicles, and monitor the flying states in order to guarantee the safety of flight tests. However, when the telemetry ground station tracks the aerial vehicle, some wild values are inevitably included in the received telemetry data due to various interference factors, which can seriously affect the interpretation of the telemetry data and the evaluation of the vehicle performance. In order to make up for the shortage of existing telemetry data wild value elimination algorithms, this paper uses the wavelet transform threshold method to eliminate the wild values in telemetry data based on the principle of wavelet transform, and introduces the swarm intelligence optimization algorithm to obtain the optimal thresholds for different telemetry data adaptively. The optimal threshold and threshold function coefficients are obtained for different telemetry data to achieve better filtering effect. Corresponding results show that the proposed method can effectively eliminate the wild values in the telemetry data and realize the filtering of the telemetry data.

As one of the main food crops in China, timely and accurate monitoring of the planting area and acreage of corn is of great significance for the evaluation of agricultural productivity and ensuring food security.Based on ESA Sentinel-2 MSI remote sensing image data, the NDVI time series curves are extracted with the support of Google Earth Engine cloud platform, the transformer model is built, and the time series data are input into the model to obtain the typical feature classification results, and the maize planting areas in typical agricultural areas of North China Plain are extracted, and the accuracy is verified by using field survey data, and compared with convolutional neural network The results were compared with those of random forest classification and convolutional neural network. The results show that the overall accuracy of transformer classification is higher and the classification effect is better compared with random forest and convolutional neural network algorithms. Therefore, the use of transformer can effectively improve the crop planting area extraction accuracy.

The reliability of data transmission in traditional protection systems mainly depends on the redundant realization of the communication network. The protection devices in the system do not have the capability and transmission mechanism for redundant data transmission. In order to solve this problem, this paper proposes a data redundancy transmission method based on protection devices, so that the protection devices in the system can be directly connected to each other and data transmission can be realized. Through the repeated transmission and reception of data by different communication ports of each protection device, a set of multi-port data dynamic transmission strategy based on protection device is established, forming a redundant transmission method that does not depend on network communication equipment. This method reduces equipment costs while improving communication efficiency and reliability.

A composite photovoltaic (PV) MPPT algorithm with improved gray wolf algorithm and perturbed observation (IGWO-P&O) algorithm is proposed, introducing a nonlinear convergence factor and Lévy flight strategy in the algorithm to optimize the overall search rate and global search capability, switching to the perturbed observation method when searching near the maximum power point (MPP), and improving the efficiency of maximum power point tracking (MPPT) of PV power generation by using its fast convergence property near the extreme value point.

Aiming at the problems of uneven distribution of ground users, reliability fluctuation of nodes and links and frequent switching of controller groups in software-defined satellite networks, a multi-controller deployment strategy for LEO satellites based on nearest neighbor propagation is proposed. The strategy aims to reduce the delay, balance the network load, improve the reliability of nodes and links and extend the effective duration of the controller group. The control domain is divided by the nearest neighbor propagation clustering algorithm and the controller group is selected. Then the simulated annealing algorithm is used to iteratively select a better performance scheme. Experiments show that the algorithm can effectively reduce the delay in the control domain, improve the link reliability, and ensure the stability of the controller group under the condition of guaranteeing the load balance of the whole network.

As 5G communication networks are now putting into commercialization, technologies for 6G communications assisted by intelligent reflecting surface (IRS) are also being explored in order to obtain faster and more reliable data transmissions. This paper studies the weighted sum-rate (WSR) maximization problem of users in an IRS-aided multiuser multiple-input single-output (MISO) downlink communication system. Aiming at the above problem, we propose a low-complexity linear alternating direction multiplier method (LADMM) that can be operated in parallel to solve this problem. The numerical results show that only adjusting the parameters in the proposed algorithm can make the user's WSR have better performance.

With the maturity of robotics, manual inspection is gradually replaced by inspection robots in large storage centres and other places where goods are densely packed. Aiming at the problems of low efficiency, high number of inflection points and inability to perform real-time obstacle avoidance in the traditional A* algorithm for mobile robot path planning, this paper proposes an improved A* search algorithm. Measures such as modification of the heuristic function, optimisation of the open list and curve smoothing are used to effectively improve the search efficiency.

Ad hoc networks are often used in scenarios such as environmental monitoring, reconnaissance reporting, and alarm. In order to realize correct and efficient data transmission of high-load nodes in the network, this paper proposes a node priority-based dynamic TDMA algorithm (NPB-DTDMA), which defines the node priority according to the usage proportion of the node's sending queue and enables the high-load nodes in the network to have advantages in obtaining information time slots. The simulation results show that when packet sending rate is lower than saturated transmission capacity, NPB-DTDMA algorithm can achieve higher throughput, lower packet loss rate, and lower average end-to-end delay than P-TDMA, E-ABROAD, and fixed TDMA algorithms.

For multipath effects and frequency fading of channels in indoor visible light communication systems. In this paper, an improved LMS algorithm is proposed to compensate for the channel. First, the channel model of indoor visible light communication and the simulation model of orthogonal frequency division multiplexing (OFDM) are built.The simulation verifies the effect of eigenvalue distribution, filter order and step size on the LMS algorithm. Secondly, the NLMS algorithm proposed in this paper solves, to a certain extent, the inherent contradiction that the convergence speed and steady-state error of the LMS algorithm cannot be reasonably coordinated. The simulation results show that: The algorithm improves the convergence speed and reduces the computational complexity compared with the LMS algorithm, which effectively improves the performance of visible light communication systems.

High-dimensional time series anomaly detection has always been an important challenge in the field of system security. Most existing methods are dedicated to modelling the temporal variation of features to capture anomalous moment points, however as features become more high-dimensional, the associations between features take on a complex spatial structure. This spatial structure information will compensate for the constraints of unsupervised training conditions, and guide the model to be more fully trained. In this study, we propose a detection model that incorporates spatial supervision signals. The model not only simultaneously models the temporal and spatial dependencies, but also simulates the topological structure and physical characteristics of data in the real world through graph structure learning and contrastive learning, providing guidance for anomaly detection. We conducted experiments on two real-world datasets and demonstrated that our model outperforms the baseline. Finally, we conducted detailed data analysis to provide interpretability for the model.

This paper presents an image dehazing algorithm based on the improved dark channel prior and heterogeneous optimization. The improved dark channel prior simplifies the window-based dark channel prior principle to pixel based dark channel prior, and employs heterogeneous acceleration techniques to improve the processing speed. To further optimize the algorithm's performance, we implement the improved dark channel prior dehazing algorithm on CPU, GPU, and FPGA. We compared the processing effects, runtime, and pixel output under unit power consumption of the dark channel prior dehazing methods based on slice bilateral filtering and guided filtering on these different types of processors. The experimental results show that the algorithm proposed in this paper runs faster on GPU and FPGA than on CPU, and also exhibits higher energy efficiency. Moreover, compared to the dark channel prior dehazing methods based on slice bilateral filtering and guided filtering, this algorithm has higher efficiency and lower power consumption while improving the image quality.

This paper implements a multi-channel data acquisition system based on AD7606 + FPGA architecture. The system can sample 8 channels of analog input data synchronously. After analog-to-digital conversion by AD7606, the whole KSZ9031 Gigabit Ethernet is packaged into LabVIEW FPGA toolkit using Socket CLIP technology, and the conversion results are transferred to LabVIEW interface using Gigabit Ethernet communication. The experimental results show that the design of this AD sampling system is effective, the sampling reliability meets the requirements and can significantly improve the control efficiency of the processor.

The dual-PWM frequency control system needs to set a filter on the grid side to suppress harmonic current. By analyzing the difference of operation characteristics between LCL filter and L filter, the parameters of L filter are optimized in order to reduce the volume of rectifier, taking into account the filtering effect, current fast tracking ability and system fast response ability; A PWM rectifier control algorithm based on L-filter is proposed, and the parameters of voltage and current loop controller are adjusted to realize the control of fast current tracking, stable voltage output and power factor of 1 on the grid side. The simulation and experimental results verify the correctness of L-filter design and the effectiveness of PWM rectifier control algorithm

This paper proposes a wind vector prediction method based on long-short term memory neural network (LSTM). The correlation between wind speed and direction is analyzed from the perspective of feature engineering. The results show that they contain different feature information and can be used as input variables to train the model at the same time. On the other hand, the above analysis also provides a basis for selecting the time length of input variables. The wind vector is decomposed into two orthogonal one-dimensional variables of east-west and north-south wind speeds based on wind direction to prevent the complexity of the algorithm from being increased by multi-dimensional variables. The LSTM algorithm is used to train the prediction model for the wind speed in both directions, and finally the wind vector prediction data containing the wind speed and direction are restored. Without increasing the complexity of the algorithm, the information density contained in the model is increased. One month's second level data of a wind farm in Hebei and Gansu provinces are selected for verification. The results show that the proposed hybrid prediction algorithm can better capture the information about wind speed and direction, and the error range of wind speed and direction prediction is reduced to 1m/s and 5° respectively, with an accuracy rate of more than 90%

With the development of various new technologies and the continuous transformation of the power system, the system safety and power quality of the new power system have become the primary focus of attention. Three-phase imbalance and harmonic, as the main factors affecting power quality, have become the focus of current research. Based on the research of traditional prediction model, this paper establishes a combined prediction model based on adaptive EEMD and LSSVM to predict the content of each harmonic in the power grid. First, the adaptive EEMD is used to separate the signals of the power grid to be measured, and the harmonics of similar frequencies are effectively separated into each corresponding IMF. Then the optimal LSSVM prediction model is established separately according to the characteristics of each IMF. Finally, the prediction results are analyzed to achieve the prediction of each harmonic content in the power grid. The experiment shows that the prediction model can effectively predict the harmonic content of different load signals in the power system with high accuracy, and the prediction effect is also relatively higher than the traditional prediction method.

In contemporary society, smart grid technology of which the foundation is substation technology plays a pivotal role in developping national economy. IEC61850 protocol is widely used in power grid technology due to its various advantages. However, many communication protocols still used in industrial fields, leading to that different devices can’t be well compatible with each other which will result in the low utilization rate. MODBUS protocol is a widely used communication protocol in the industrial scope. This paper mainly study the MODBUS RTU and IEC61805, and then design the transformation of the two protocols by means of object-oriented modeling, as well as design Restful API interface to collect data for visualization. Later, this method of conversion can also be applied to the conversion between other protocols, such as DNP3 and CAN protocols, which will accelerate the pace of protocol unification and promote the development of substation automation.

This paper makes use of the powerful interface development ability of Visual C# and the numerical operation ability of Matlab, and uses the mixed programming technology of Matlab and Visual C# for software programming design. First, create the PeakWave.dll of the peak radiated noise samples and the ShoulderWave.dll of the shoulder noise samples, then call the corresponding functions in the main program, and finally return the noise samples generated by simulation by MATLAB, thus realizing the design of the ship radiated noise simulation software with simple operation and good user interface. The software has the functions of noise simulation, parameter setting, data saving, waveform display, etc. The results show that the characteristics of Visual C# and Matlab can be brought into full play by using mixed programming technology, and the software designed has a wide range of engineering application value.

In order to effectively solve the problems of power system front-end technology diversification and low explicit programming efficiency, a "visual programming" method based on low-code development platform is proposed. This method uses component-based graphical programming tools to build front-end pages, designs data collection forms through the platform, and delivers the collected data to the back-end neural network model for training, so as to predict the fault risk of distribution line units, improve the power enterprise's ability to adapt to demand changes, carry out operation and maintenance more pertinently, and greatly improve the business staff's ability.

Remote sensing image classification has experienced three stages: pixel-level, object-level and scene-level. With the improvement of remote sensing image resolution, pixel-level and object-level methods cannot be completely correctly classified, and thus, scene classification is the current focus of this research. We consider the complex background of remote sensing images, the existence of many small objects and the large scale of change, as well as intraclass diversity and interclass similarity. Through the salient regions and features in remote sensing images, a dual attention dense network is proposed. In addition, an adaptive spatial attention module and an adaptive channel attention module are designed. Specifically, the network combines the output of the two proposed attention modules as the feature representation. Among them, the adaptive parameter activation function is introduced into the adaptive spatial attention module, and different nonlinear transformations are performed on the input features in the spatial attention network to achieve attention on important regions. By capturing the adaptive cross channel interaction range to learn channel attention, important weights of each channel are generated and an adaptive parameter activation function is introduced to adjust the feature values of different channels, thereby acting with the global features to achieve attention on the salient features. We present extensive experiments on three scene classification datasets, including the UCM dataset, the AID dataset and the OPTIMAL dataset, and compare them with various algorithms. The experimental results demonstrate the effectiveness of our proposed dual attention model.

The nonlinear characteristics of the passive devices will cause the nonlinear distortion of the system. Passive intermodulation is a typical nonlinear interference phenomenon. Coaxial connector as a widely used passive component will cause passive intermodulation, which is attributed to material nonlinearity and contact nonlinearity, and the environmental temperature will affect the contact characteristics and material characteristics of connectors. In this paper, considering the contact nonlinearity, the relation model between passive intermodulation performance and temperature was established theoretically, and the accuracy of the model was verified by a series of heating degradation experiments, and a real-time test system was designed to evaluate the passive intermodulation performance in the process of temperature cycle.

This study is devoted to a description of a loop closure detection framework, in which the leveraging of a VGGNet-19 and a K-means cluster enables a practical, autonomous feature learning-based detecting. The principal components analysis (PCA) for dimension reduction is also investigated, guaranteeing the algorithm optimization in both accuracy and efficiency. In terms of benchmark dataset tests, the results are compared against bag-of-words (BoW) model, AlexNet and VGGNet-16, revealing our proposed design significantly outperforms others in Precision-Recall. The calculated cosine similarities and the detected closed-loop frames are simultaneously provided.

The task of Knowledge Base Question Answering (KBQA) is to answer a question in natural language over a Knowledge Base. And multi-hop KBQA aims to reason over multiple hops of facts in KB to answer a complex question. Step-wised reasoning has been an important schema to solve multi-hop KBQA. But previous approaches suffer from lacking reasoning paths, causing models may answer in an incorrect way. To address the issue, we present a novel approach to enhance the KBQA model by leveraging consistency between different views of the data, with few intermediate-relation-labeled data. Previous retrieval-based methods proceeded by utilizing the data view of (question, intermediate entities, answer entities). In our method, we introduce the data view of (question, intermediate relations) and enhance the KBQA model through the consistency of different data views. Concretely, we first implement a question-to-intermediate relations(Q2R) model to obtain intermediate relations’ distributions. By utilizing a pretrained text generation model, it performs well using a small part of relation-labeled data. Then we devise a map function to map distributions of intermediate entities to distributions of intermediate. Finally, a constraint that metrics the consistency between the intermediate path distributions obtained from the Q2R model and the original KBQA model is constructed to enhance the KBQA model. Experiments over three datasets of multi-hop KBQA are conducted, and the results demonstrate the effectiveness of our method.

Short-term power load forecasting plays an important role in power system dispatching. To improve forecasting accuracy, a short-term load forecasting model based on stacking ensemble learning was proposed. Firstly, add effective multi-feature variables, and establishes a Stacking ensemble learning model for the load data and feature, which was ensembles by Light Gradient Boosting Machine (abbr. LightGBM) and eXtreme Gradient Boosting (abbr. XGBoost) for prediction. Finally, the comparison and experimental results show that the forecasting error of the proposed model is less than that of the comparative model.

In this paper, a portable EEG detection hardware system based on ADS1299 is designed. The hardware system mainly integrates the analog front-end chip ADS1299 to collect EEG signals and the STM32 main control chip to process EEG signals, and then transmit the data to the host computer through the integrated Bluetooth low energy chip, and use the host computer to process and analyze the EEG signals. The volume of the system meets the size requirements of portable EEG acquisition devices on the market, and can be directly connected to dry contact points, unlike traditional medical wet sensors that require conductive glue. With 8-channel acquisition and processing capability, low cost and low power consumption, it is suitable for portable consumer products and other applications.

This article explored the effect of physical activity time and that of the heart rate during physical activity on mental fatigue intervention. First, we acquired electrocardiogram from 104 subjects, calculated the inter-beat intervals (IBI), removed the abnormal values in the IBI series, segmented and divided the IBI series into training and validation datasets and application datasets according to the inclusion criteria of mental fatigued and non-fatigued states. Second, we extracted 39 linear and non-linear RR parameters as fatigue physiological features, and applied Leave-One-Subject-Out cross test and Sequential Backward Selection algorithm for feature selection while training some traditional classifiers. Third, we applied the best trained classifier to detect mental fatigue before and after physical activity. Finally, we analyzed the change of mental fatigue status before and after physical activity and obtained the following three findings: (1) physical activity did not have intervention effect on mental fatigue for 62% observed cases; (2) for 38% observed cases, physical activity could aggravate or relieve mental fatigue; (3) for the observed cases that the physical activity time was in the range of 5-15 minutes, the average heart rate in 100-140 beat per minute during physical activity was more likely to relieve mental fatigue

The unmanned underwater vehicle (UUV) based on waypoint navigation is used as the research object to optimize the multi-objective trajectory for the smooth operation, low energy consumption and low smooth impact required in the patrol task. The spatial trajectory of the unmanned underwater vehicle is constructed by a quintuple polynomial, and the motion trajectory is optimally solved using a quadratic programming algorithm by combining the position, velocity and acceleration requirements of the unmanned underwater vehicle at the beginning and end moments as well as the continuity constraints among the waypoints. The results show that the multi-objective quintuple polynomial-based algorithm achieves an effective multi-objective optimization of the unmanned underwater vehicle trajectory.

The action evaluation task is to automatically analyze and evaluate the actions made by the human body. Push-up action is one of the physical fitness test items. At present, the traditional manual evaluation method is mainly used for its evaluation. This method is inefficient and the results have certain subjectivity. Based on the "National Physique Measurement Standards", this paper selects the joint points describing the push-up movement, detects the data of the push-up joint points based on Kinect, and calculates the 8 joint angles of the push-up as its action characteristics through the vector method, and then proposes based on the dynamic time warping algorithm A new evaluation method is developed, and finally a push-up evaluation system is designed and implemented.

Duo to the wide band gaps, fast saturated electron drift velocities and high breakdown electric field strength of the silicon carbide (SiC), it is an appropriate candidate to develop power supplies working in high temperature and high voltage environments. Usually, the requirement on the reliability of these devices is much higher than those of the universal power supplies. Based on the real device structure of the 4H-SiC MOSFET, the output characteristic is simulated with TCAD package and verified by comparing with the testing results from the datasheet, which provides the data set for training BP neural network. Furthermore, an BP neural network is trained to predict the output characteristics of the MOSFET. Agreement between the predicted characteristics and real characteristics is achieved. The trained neural network can be easily integrated in embedded system and provides the possibility for health monitoring and fault diagnosis based on artificial intelligence.

When the frequency of power signal is offset, it is difficult to achieve synchronous sampling by using discrete Fourier transform in synchronous phasor measurement. The fence effect caused by this will seriously affect the precision of synchronous phasor measurement, especially the precision of phase measurement. Therefore, this paper proposes an improved DFT synchronous phasor measurement algorithm. Firstly, an algorithm for frequency tracking, which is based upon the extended Kalman filter, is put forward. Besides, the algorithm for particle swarm optimization is described with a view to optimizing the noise weight of Kalman gain. Based on the frequency offset rate obtained by frequency tracking, the improved DFT algorithm is adopted for the sake of attaining the corrected phase result. These consequences of simulation indicate that the frequency tracking error of this algorithm doesn`t reach 0.015Hz, and the phase calculation error doesn`t reach 0.018°, which satisfies the accuracy engineering demands, and the convergence speed is fast.

THz metasurfaces that reflect radiation back can be applied in various fields such as imaging, biosensing, and optical communications. However, the conventional THz metasurfaces have limitations due to their inflexible electromagnetic responses and complex structures. In this paper, a voltage-controlled dual-polarization metasurface based on vanadium dioxide (VO2) is proposed, which can achieve the polarization conversion and transmission by controlling the conductivity of VO2 in the surface. The results show that when the VO2 is in the metal state, the metasurface can realize the circular polarization mode conversion in the range of 1.2-1.4 THz. While insulated, it will switch to the low-efficiency transmission conversion mode. Then, A broadband voltage-controlled OAM beam generator metasurface is designed, providing a method for realizing high performance multifunctional tunable metasurfaces in the THz band. This work has the potential to expand the practical applications of metasurfaces.

In recent years, the use of rooftop photovoltaic (PV) has increased as countries upgrade their energy systems. However, estimating the impact of behind-the-meter PV on grid operation requires precise physical models and weather information, which is not practical. To address this issue, we propose a data-driven approach using a sequence-to-subsequence (Seq2subseq) PV decomposition model based on the one-dimensional convolutional neural network (1D-CNN). This model automatically extracts temporal features from net metered sequences and outputs behind-the-meter PV generation using a sliding window. We evaluated our model on 184 rooftop PV users in the SGSC dataset, demonstrating its accuracy and ability to generalize across different climates. Our proposed approach offers an effective solution for real-world applications.

A low-power RISC-V-based convolutional neural network acceleration processor is proposed to cope with the problem that the increasing resource requirements of convolutional neural networks in the direction of hardware convolutional acceleration are difficult to be met on embedded devices. The processor is designed with three instructions that can configure the parameters of each CNN layer to accommodate different input data, multiplex computational resources to reduce power consumption, and execute operations that repeat a large number of executions in parallel to speed up operation efficiency. Through comparison experiments, it can be found that this processor acceleration instruction set is 20.93 times, 7.67 times, and 8.97 times faster than the base RISC-V instruction set after verified with the same data on three operations, including convolution, activation, and pooling, respectively. The experimental results show that the total power consumption of the processor with this custom instruction set is only 0.221 W at 16 MHZ operating frequency, which is advantageous in terms of performance-to-power ratio compared to other RISC-V accelerated processors with less resource consumption and lower power consumption.

In order to quickly and accurately identify insulator defects and solve the problems such as labor waste and misjudgment of traditional detection methods, this paper proposes an insulator detection algorithm based on YOLOv3(You Only Look Once). Firstly, the K-means algorithm is improved by the genetic algorithm to generate anchors, then the YOLOv3algorithm is trained using the anchors, and finally the trained model is used to detect the insulator images. After experimental comparison, the algorithm in this paper is more accurate than the original K-means algorithm to generate anchors, and the YOLOv3defect detection model is more accurate than the original algorithm.

With the increasing automation of aircraft ground assembly integration testing, the progress of aircraft ground function testing is severely impacted by the multitude of devices, types, and interfaces present at civil aircraft final assembly sites. Some devices even lack communication interfaces. Embedded system-based interlayer cooperative interaction is a general trend and the future development direction. It is urgent to enable data interaction between the hardware and software layers using advanced embedded microprocessor technology. To address this challenge, this paper proposes a Mirco XRCE-DDS (EXtremely Resource Constrained Environments DDS) solution based on eProsima that adheres to the OMG (Object Management Group) specification and is suitable for resource-limited equipment, to facilitate the design of civil aircraft ground function testing systems

With the change of observation demands, traditional monostatic synthetic aperture radar(SAR) can no longer meet the requirements such as passive reception and cooperative multimode imaging. Spaceborne/Airborne hybrid bistatic SAR(SA-BSAR) emerges as the times require, and has been widely concerned with its flexibility, concealment. However, as the receiver and transmitter do not share the same frequency oscillation source, the impact of time error on imaging cannot be ignored as the satellite operates for a long time. At the same time, SA-BSAR is in the form of double root sign in the course of slant range. How to obtain two-dimensional(2-D) spectrum is also the key to imaging. This paper proposes a new method to estimate time error based on fractional Fourier transform (FRFT), which simplifies the traditional time error estimation steps. Finally, the Air-Phase method is used to solve the 2-D spectrum to achieve the target imaging. The algorithms proposed in this paper are implemented by FFT without interpolation, which ensures the imaging efficiency. The final experimental results show the effectiveness of the algorithm.

The current communication network can complete communication, entertainment, shopping and other applications, in which mobile phones play an important role, and mobile phone communication and Internet access need to rely on base stations. This paper proposes the optimization of communication network from the perspective of network structure, analyzes the factors affecting the communication network, and optimizes the establishment and deployment of communication base stations. On the one hand, it provides customers with better communication quality and faster communication speed, and on the other hand, it improves the investment income of communication operators.

In order to improve the efficiency and decision-making ability of airport operation support, the realization of estimation about service time of flight ground support can reduce the time and economic losses caused by flight delays. Considering the complexity and particularity of the service process, this article started from the analysis of the flight ground support process and constructed a mathematical model of the service time. The method of Principal Component Analysis (PCA) was used to reduce the correlation between variables, and a service time prediction model of flight ground support based on Deep Neural Network (DNN) was established. Finally, the flight support operation data of an airport were selected for simulation and verification. Experimental results show that the average absolute error of service time prediction can reach 2.709 min, the proposed model can effectively estimate the service time of flight support and has higher accuracy.

Many biomedical ontologies develop regularly and change over time. An ontology new release will update its data, containing that fix some errors in the previous version and add many new concepts to adapt to the development in the domain. Insertion of new concepts into their proper positions on a terminology is a challenging problem in the automatic enrichment of ontologies. In the past, the new concepts are always created by domain experts. Then the experts will run a traditional classifier or manual operation to insert the new concepts in proper place. With the development of technology, the methods based on Machine learning (ML) have been proposed to help terminology researchers to develop and maintain the ontologies. We propose an new approach that is based on providing only the concept name and using a Graph Convolutional Network (GCN) aggregated the sub-string neighbor information learning method. We chose a Bidirectional Long Short-term Memory Networks (Bi-LSTM) model as our classifier for the predicted task. We first tested this method within Gene Ontology (GO) 2020 January release and achieved an average of 89.68% precision and an F1 score of 0.9081 in task of predicting direct IS-A links. In comparing the January 2020 release with the March 2022 release, we predicted the links related to new concepts, our average Accuracy score was 0.6996.

In order to solve the problem of network real-time data transmission and the difficult requirements of monitoring system for network data real-time presentation, this paper designs a power monitoring configuration software with Web service function. It is based on B/S architecture, which can expand the equipment primitives for power monitoring applications, and has strong universality and transplant reliability. By using its configuration function, users can quickly generate the power monitoring software system, which can realize the scientific management of power grid monitoring in power enterprises and achieve the purpose of improving work efficiency and automation management level.

In recent years, the speed and capacity of fiber optic communication systems have been greatly improved, but in the event of a failure requiring a switching, the 50ms switching time defined by the International Telecommunication Union Telecommunication Standards Branch (ITU-T) can cause a large amount of service data loss compared to low-speed transmission. Therefore, how to minimize the switching time has also become a key issue in protecting the switching. In this paper, we use a magneto-optical switch (MO-Switch) as the core device for protection switching, and design a state machine using FPGA/CPLD to control and design an automatic protection switching (APS) protocol in hardware to realize a 1+1 unidirectional or bidirectional linear protection switching system. The proposed scheme finally achieves a switching time of about 1.5ms, which achieves a performance improvement of more than 95% compared to the standard 50ms and reduces the loss caused by the failure of high-speed fiber optic communication systems.

To achieve that the mobile warehouse robot follows the given desired path quickly and smoothly, the MPC and LQR steering control algorithms are applied based on the lateral kinematic constraints of the vehicle. First, the Ackermann kinematic modelling of the mobile platform is performed. The nonlinear model is linearized and discretized to create a discrete state space model of the mobile robot. Under the same conditions, a lateral control system based on MPC and LQR is designed for the mobile robot. A performance comparison of parameters such as different vehicle speeds, straightline trajectory tracking, curve trajectory tracking and algorithm consumption time is performed. The simulation shows that the LQR and MPC controllers can calculate the vehicle's steering angle in real time according to the road curvature and drive according to the preset desired path.

Most news recommendation methods believe that all news in the implicit feedback sequence is beneficial to the final result, but it includes the influence of irrelevant news (i.e. false user behavior). This is not ideal for mining user interest information, which usually contains large noise. In order to address this issue, a news recommendation model based on positive and negative implicit feedback feature fusion with sparse attention mechanism is proposed. Firstly, the method adopts the method based on deep learning. In the user interest representation module, the adaptive sparse transformation function is introduced to eliminate the influence of irrelevant news in the user's positive and negative implicit feedback sequence. Secondly, the user's interest features are better constructed by fusing the user's positive and negative implicit feedback features. Finally, the proposed model is examined and analyzed on the MIND dataset. Compared with other news recommendation methods, the recommendation evaluation index of this method has been effectively improved.

Sepsis is a complex syndrome that leads to shock or death in the intensive care unit (ICU). The influence of multiple factors combined with the patient’s heterogeneity makes treating sepsis challenging to personalize. Research in reinforcement learning (RL) has made significant progress in medical decision-making in recent years. In addition, combining deep learning with RL is gradually becoming a mainstream approach to solving complex sequential problems, which provides a solid theoretical foundation for our work. This work proposes a supervised reinforcement learning (SRL) model to recommend personalized fluid and boosting drug doses for sepsis patients. In this context, RL aims to maximize the expected return to find the optimal treatment strategy, and supervised learning minimizes the difference with physician decisions to reduce the occurrence of high-risk strategies. The two complement each other, allowing the model to learn reasonable treatment strategies to assist physicians in making medical decisions to improve treatment outcomes. The results show that the SRL model can discover more optimal treatment strategies to provide personalized interventions.

As the security issues of Internet of Things (IoTs) are rapidly evolving, machine learning techniques are increasingly adopted for detecting and preventing cyber threats. Recent machine learning based approaches (e.g., anomaly detection, intrusion detection, and predictive analytics) are being utilized in IoTs security. With the proliferation of IoTs devices, it is crucial to develop scalable and effective security solutions to keep pace with the changing threat landscape. This paper proposes a novel NSM (Network Sparsification Modeling) approach for identifying and categorizing cybersecurity threats in the cloud and IoTs environment. The proposed NSM algorithm is to optimize the Kullback-Leilber divergence based on higher-order spanning k-tree modeling process. The NSM model is capable of detecting cybersecurity threats in the cloud and IoTs setting by converting raw data into a meaningful format. The performance of the NSM model was evaluated using CICIDS 2017 dataset. The testing results prove that NSM model is state-of-the-art by outperforming others. Future deep-learning approaches are capable to integrate in the ML-based NSM model for further enhancement.

Conventional algorithms typically rely on system identification techniques to estimate the inertia of power systems online. However, selecting an appropriate model order can be challenging, and an incorrect choice can lead to significant errors. To address this issue, we propose an algorithm based on Long Short-Term Memory Network (LSTM) deep learning networks for power system inertia identification. In our approach, we preprocess and input frequency and power deviation data obtained from monitoring into the LSTM model for learning. Additionally, we utilize the multi-sampling point method to reduce errors introduced by approximation algorithms. Once we obtain the inertia time constant for each unit, we calculate the system's overall inertia. Finally, we build a simulation system using MATLAB/Simulink to demonstrate the effectiveness and accuracy of our proposed method.

This article introduces deep learning into the multiple-input multiple-output (MIMO) sparse code multiple access (SCMA) system and proposes a MIMO-SCMA detection scheme based on deep neural networks (DNN) to improve bit error rate (BER) performance. The DNN learns the codebook of each user through channel feature learning on different transmission antennas. The fully connected DNN is designed as the decoder at the receiving end, which does not require traditional multi-antenna detection and multi-user detection, and can obtain user data with one decoding operation. The encoder and decoder are trained using an end-to-end training method. All learning models of the DNN are generated offline and the learned models are used for online testing. In this model, the received signal and channel coefficients are set as input data, and the label corresponding to the transmitted symbol is set as output data for offline learning. After offline learning is completed, the model can be deployed online with fixed weights and biases. Through simulation experiments, the proposed DNN encoder-decoder method can reduce the BER and computational complexity of the receiver in the MIMO-SCMA system.

The electrostatic precipitators are widely used in power plants and other energy fields.The invention discloses a device for on-line processing short circuit fault of the electrostatic precipitator electric field without stopping, which comprises the robot installed on the top shell platform of the electrostatic precipitator, and the robot is controlled and connected with the monitoring device outside the electric field, so that the monitoring device can remotely control the robot wirelessly moving inside the electric field of electrostatic precipitator to the short-circuit fault point of the electric field. The device can effectively solve the serious problem of electric field short circuit caused by cathode line fracture or hypertrophy and the problems of time-consuming, laborious, high cost and serious pollution, so as to achieve the purpose of cost saving, quality improvement and efficiency.

Aiming at the problem that the current power OPGW optical cable early warning model adopts a single technical index, this article proposes an early-warning model of power OPGW cable operating status based on joint judgment. Firstly, a multi-element joint judgment power OPGW optical cable prediction model is constructed based on the Long Short-Term Memory (LSTM) model, and the predicted power OPGW optical cable core strain sequence provides a data basis for the early warning of the operation status of the optical cable. Then, based on the identification and classification results of the Multilayer Perceptron (MLP) model, the early warning level is divided, and the core strain data of the power OPGW optical cable obtained by the prediction module is identified and output the early warning level, so as to realize the early warning of the operation status of the power OPGW optical cable. Finally, model experimental analysis is carried out. The experimental results show that the proposed model can provide early warning for power OPGW optical cable, and further improve the risk management and control strategy of OPGW optical cable, so as to complete the operation and maintenance of power OPGW optical cable more efficiently

In this paper, we study the path planning problem of a fixed-wing unmanned aerial vehicle (UAV) with a reconnaissance camera when performing a reconnaissance mission in airspace containing a no-fly zone. Due to the UAV's dynamic constraints and the reconnaissance mission's specificity, the problem can be formulated as a special variant of the traveling salesman problem (TSP), called the dynamic-constrained TSP with neighborhoods (DCTSPN). To solve this problem, the authors propose a hierarchical algorithm based on deep reinforcement learning (DRL), divided into an optimal sequence planning layer and a shortest path planning layer. In the optimal sequence planning layer, Firstly, the neighborhood boundary of the target point is discretized to form multiple reconnaissance points; then, the complex trajectory planning problem is simplified to planning on a finite directed graph by randomly selecting a finite set of reconnaissance points from the neighborhood boundary of the target point set. The double deep Q network (DDQN) algorithm is used to solve for the target point traversal sequence and the reconnaissance points for each target. In the shortest path planning layer, Use the Deep Deterministic Policy Gradient (DDPG) algorithm to perform global path planning in continuous state space and action space and generate a dynamically feasible optimal flight trajectory under the guidance of the optimal reconnaissance sequence to complete the given reconnaissance mission, Avoid no-fly zones. The simulation results show that the hierarchical algorithm is highly applicable and efficient.

In this paper, a cable aging state assessment method based on the combination of XGBoost and FocalLoss was proposed. Firstly, the FocalLoss function is used to deal with the problem of small cable sample data and serious imbalance in the ratio between classes. Secondly, the FocalLoss function is used as a custom loss function in the XGBoost algorithm, and the combination of the two can achieve effective assessment of cable aging status by extracting key features of cables. Finally, it is verified by example that the evaluation method can effectively deal with the problem of small number of samples and imbalance, and the accuracy of cable aging state evaluation is significantly improved, which can provide a new direction for cable aging state evaluation.

An improved K-means clustering algorithm based on the initial clustering center is used to cluster the charging data of electric vehicles. The multi-classification method is studied, and the clustering effect of different number of clusters is analyzed with the contour coefficient as the evaluation standard. The simulation results show that this method can properly cluster the group characteristics of electric vehicles

In this paper, a unified model of island division and network reconstruction of active distribution network including distributed power supply and energy storage is constructed, and the fault maintenance strategy and load demand response strategy are considered to realize the integrated operation of "source-network-load-storage". Then the second order cone relaxation technique is used for convex relaxation, and the original model is transformed into a standard mixed integer second order cone model, so as to reduce the complexity of the solution. Finally, IEEE 33 node active distribution network is used to verify the superiority of the proposed fault recovery strategy.

In order to further improve the Internet of Things technology and ensure the ability and effect of the on-line monitoring system for electric energy metering. A parameter monitoring method of electric energy metering system based on Internet of Things technology is proposed. In this paper, the necessity and composition of the transmission line online monitoring system based on the Internet of Things technology are summarized. The sensing layer carries out real-time data exchange by using wireless communication interfaces and controlled devices, or supervises the components of intelligent sensors in a consistent way, and collects the information effectively, and classifies the real-time data into the controller. The network layer communication equipment is used to accurately transmit the collected information to the electric power network system, and it provides further reference for the management of various equipment scheduling. By using different information transmission methods, the unified information can transmit the perceived data to the circuit system, and the data can be classified and processed in a unified way. And all the effective information is collected in the controller for real-time data forwarding; The most important function of the application layer is to further process the information from the network layer, and at the same time, store it in a unified code, and then judge whether there are some faults in the line according to the information. At the same time, the problems existing in the running circuit are concretely speculated by comparing the values of normal operation and variables. With the intelligent development of on-line monitoring system for transmission lines, we need to further expand various technologies such as state sensors, and further establish and improve the intelligent monitoring model for transmission and transformation equipment.

With the rapid development of new power systems and the stringent communication requirements of electric hybrid services, the 5G interface and access control technologies are required to improve to meet the differentiated demands of low latency, massive connection, and so on. In this paper, we study the 5G interface and access control technology for massive device connection in electric hybrid service scenario. The key 5G interfaces and their functions are firstly introduced. Then, we consider an average access delay minimization problem under the minimum signal-to-noise ratio (SNR) and the maximum delay requirements. A reinforcement learning-based access control algorithm is proposed to learn and dynamically adjust the optimal base station selection strategy to alleviate access congestion and reduce the access delay. Simulation results show the superior performance of the proposed algorithm in the average access delay and collision probability

With the development of new electric system, the new multi-scenario electric services put forward more stringent and differentiated communication requirements. 5G communication with software-defined network (SDN) and network functions virtualization (NFV) technologies provides an efficient solution. In this paper, we firstly analyze the communication requirements of multi-scenario electric 5G services including control, acquisition and mobile application services. Then, we propose a SDN and NFV based slice management scheme of electric 5G communication network, which aims to minimize the total transmission delay by joint optimizing transmission power and network slice allocation under the delay and energy consumption constraints. Simulation results show the superior performance of the proposed scheme in convergence rate, total transmission delay, and differentiated communication demand guarantee for multi-scenario electric 5G services.

This paper, originally published on 11 August 2023, was withdrawn on 10 October 2023 per author request.

Based on the advantages of Distributed Feedback fiber laser (DFB-FL), such as narrow line width, low noise and long coherence length, a vibration sensing system was designed and used to detect transformers vibration in this paper. The DFB-FL glued on the cantilever was used as the sensing element, the wavelength change of DFB laser was converted into the output intensity change of the interferometer by using unbalanced fiber Michelson interferometer. Cantilever was designed and vibration characteristics was simulated by COMOSOL software with different material. The vibration signal of the winding of the transformer was detected with vibration frequency centered at 100Hz,750Hz, 800Hz and 850Hz.

With the increasingly prominent environmental problems and the promotion of the “double carbon” goal, the participation of demand-side response resources in load peak shaving and valley filling is of great significance to the optimal operation of the power system. Based on the background that China's power demand-side response is rapidly developing and helping to implement the carbon peak target, this paper starts with optimizing dispatching and replaces the traditional price target with the system carbon emission target. Firstly, the stochastic fluctuation model of wind power and photovoltaic output is added to the power dispatching model of thermal power units, and the dispatching model taking into account the fluctuation of wind and solar power is established. Secondly, the uncertainty model of demand response is introduced, and the optimal dispatching model of power system considering the uncertainty of source and load is established. Then compare and analyze the dispatching models based on carbon target and cost target, and analyze the advantages and disadvantages of demand response in smoothing load, reducing carbon emissions and reducing power shortage risk after participating in power dispatching under different targets. Finally, through the analysis of the scheduling results of a typical peak-shaving scenario in a region, the important role of demand response resources in reducing the peak-valley difference of the power system and reducing carbon emissions is verified.

Convolutional neural networks (CNNs) have been widely used in the field of image recognition. To meet the massive computational requirements of CNNs, GPUs or other intelligent computing hardware are typically used for data processing. FPGA supports parallel computing and is characterized by programmability, high performance, low energy consumption, and strong stability. In this paper, we improved and optimized the YOLOv2-Tiny algorithm by combining it with the hardware implementation based on FPGA's hardware structure. We divided the neural network tasks and preprocessed data using the 16-bit fixed-point method to reduce hardware resource consumption. By using the PYNQ-z2 development platform to accelerate the YOLOv2-Tiny CNN, we achieved target object detection and recognition. Compared with CPU (i7-10710U), the processing capacity was 2.94 times that of CPU, and the power consumption was 3.1% of CPU.

In the future, with the large-scale integration of distributed generation (DG) and electric vehicle (EV), due to the dual uncertainty of time and space, it is bound to pose new challenges to the economic and safe operation of urban distribution network. As one of the important means of power grid optimization, distribution network reconfiguration can dynamically adjust the power grid structure according to the spatial and temporal changes of EV charging load. Therefore, in order to improve the economy and safety of urban distribution network operation, this paper proposes a dynamic reconfiguration model of active distribution network considering EV charging demand under the guidance of real-time electricity price. At the same time, the reconfiguration period is divided based on the peak-valley membership degree. The ratio of active network loss at each moment of the system and the operating loss cost after the introduction of time-of-use electricity price is used as the operation index, and the reconstruction period is reasonably divided by the change rate of membership degree. The demand response (DR) mechanism is introduced before the reconfiguration, and the active distribution network reconfiguration model with the minimum operating loss cost is established. The model is solved by the improved binary particle swarm optimization algorithm. Finally, a case study of a city's traffic network and an improved IEEE33 node coupling system is carried out to verify that the time-sharing reconstruction method in this paper can effectively deal with the influence of DG output, EV charging and other factors on the urban distribution network, and improve the economy and safety of the overall distribution network operation.

It is particularly important to improve specific energy of the power system, lightweight design of battery box before battery technology has made a major breakthrough. The paper uses Hypermesh software to analyze and solve the multiple working conditions of battery upper box. According to analysis results, OptiStruct software was used to optimize the shape of upper box. Through the reasonable arrangement of stiffeners, the mass of battery upper box was reduced by 12.1%, and the first-order natural frequency was increased by 5% to meeting the design requirements. This optimization scheme has certain reference significance for the design of battery upper box.

With the increasing reliability of power supply in distribution networks, loop closing operation is often required to achieve non-stop load transfer. Firstly, the loop closing device based on improved phase shifting transformer (IPST) can be effectively used in complex loop closing scenarios such as three-phase asymmetry and voltage magnitude difference between the buses on both sides of the loop closing point of the distribution network, considering that the internal impedance of the loop closing device has a certain influence on the quality of the loop transfer voltage. Secondly, the mathematical model of IPST equivalent impedance is established based on the basic principle of transformer, and the expressions of IPST equivalent impedance calculation and the factors affecting the size are derived and analyzed. Finally, the correctness of the proposed IPST-based impedance calculation expression for the loop closing device is verified by MATLAB and PSCAD/EMTDC simulation platform.

In this paper, the optimal configuration of cooling, heating and power energy storage (CHPES) under the typical energy system architecture of commercial buildings is studied. On the basis of meeting the user's demand for electricity, heating and cooling, the configuration capacity and power of electric, heating and cooling energy storage are optimized with the goal of minimizing the total cost of equal annual value within the life cycle of the system. By dividing typical days, the variable dimension is greatly reduced and the solving time is shortened, so that the optimization model can be solved by the commercial solver. The numerical example shows that, CHPES can significantly reduce the equal annual cost of the system, and also promote the peak shaving and valley filling of the power grid.

For the purpose of the present study of lithium battery SOC estimation, fractional-order calculus theory and the fact that the real capacitance is fractional-order in nature mean that integer-order modeling yields incorrect methods. To improve the accuracy of lithium battery state-of-charge (SOC) estimation, a fractional-order traceless Kalman filter technique is proposed with a second-order RC fractional-order model, and a least-squares approach with a variable forgetting factor is utilized to determine battery parameters. The system gives real-time updates to the battery condition and settings through recursive estimation of state and parameter variables. Simulation analysis is performed using experimental data and UDDS operating parameters. The traceless Kalman filter method's simulated values are compared to the simulation outcomes. These results show that the method beats the integer-order traceless Kalman algorithm and that the maximum estimation error of battery SOC can be maintained below 2%. This proves that the proposed approach works as intended.

This paper, originally published on 11 August 2023, was withdrawn on 11 October 2023 per author request.

With the rapid development of UHV DC transmission projects, a large number of transmission lines need to cross high altitude areas. The reduction of corona inception field strength in high altitude areas makes corona discharge more intense. The resulting electromagnetic environment problems such as audible noise and radio interference are one of the key factors restricting the development of UHVDC transmission projects. In order to study the influence of altitude on the corona effect of negative conductor, four different altitude points in the range of 1120m-4320m were tested by using mobile small corona cage. The current pulse, audible noise and radio interference level of 0.8mm diameter fine copper conductor were collected. Through the analysis of the experimental data, the following conclusions are obtained: with the increase of altitude, the amplitude and number of current pulses increase. The audible noise and radio interference levels increase with the increase of altitude. The conclusions obtained in this paper have certain reference significance for the control of electromagnetic environment of UHVDC transmission project.

In order to further improve the separation and detection accuracy of bearing fault characteristics, A new method for early fault diagnosis of rolling bearings based on Maximum Correlated Kurtosis Deconvolution and autocorrelation kurtograph was proposed. Firstly, the vibration signal of bearing fault is denoised by Maximum Correlated Kurtosis Deconvolution; Then, the improved autocorrelation spectral kurtograph is used to select the optimal frequency center and bandwidth of fault features. According to the optimal frequency center and bandwidth, the band pass filtering is carried out to remove noise and random pulse irrelevant components in the band signal. Finally, the sub-signal after bandpass filtering is analyzed by envelope spectrum, identify fault frequency and realize early fault diagnosis of rolling bearing. In the experiment, different types of bearing fault data verify the effectiveness of the proposed method.

This paper proposes a self-attention LSTM (SALSTM) model for ship motion prediction, which combines the advantages of LSTM and self-attention mechanisms. The model also introduces the concept of attention gate. The paper studies the influence of forecast lead time on the prediction accuracy of three degrees of freedom: roll, surge and heave. The paper compares the SALSTM model with a baseline LSTM model on a ship motion data set under different forecast durations and lead times. The paper evaluates the performance of the SALSTM model using four metrics and verifies its effectiveness under three representative working conditions. The paper also gives the applicable conditions of the SALSTM model for ship motion prediction

This paper, originally published on 11 August 2023, was withdrawn on 12 September 2023 per author request.

The contact resistance between the pantograph slide and the contact wire is a key parameter to characterize the current quality of the pantograph-catenary system. In order to analyse the influence of contact pressure, sliding speed, current, and physical performance parameters of the slide on the contact resistance under high current and high speed. The contact resistance simulation model of pantograph-catenary system was established by using the COMSOL Multiphysics software, the validity of the model is verified by experiments and the simulation results show that under the simulation conditions of high speed and high current, the contact resistance decreases with the increase of contact pressure and current, and increases linearly with the increase of sliding speed. Under the same simulation conditions, the contact resistance decreases with the increase of conductivity and thermal conductivity of the pantograph slide. When the hardness of the slide increases, the contact resistance shows a trend of rapid increase first and then slowdown.

Weather radar simulation technology has many applications in civil aviation, military and other fields. Using the emulator, weather radar signals from the past can be observed repeatedly. With the development of electronic circuit technology, weather radar emulators develop to a faster response speed. This paper introduces a weather radar emulator based on DDS technology and details its design and realization. The test results show that the system can output the waveform specified by the upper computer and drive the upper computer to produce weather radar images.

With the increase of the share of wind power in energy distribution, accurate ultra-short term wind power prediction results play key role in the optimal real-time scheduling of the power grid. A stacking integration method is proposed based on error correction in this paper. First, the support vector machine for regression (SVR), gradient boosting decision tree (GBDT), multilayer perceptron (MLP) and random forest (RF) are selected as the base models. Then, the linear regression is utilized as the meta-model. The error generated by the base model in the verification set and the spliced verification set are introduced into the training set of the meta-model. Finally, the prediction results and prediction errors in the prediction set are applied to the meta-model to predict the ultra-short term wind power. The experiment results show that the effectiveness of the proposed method by using the real wind power data.

To solve many problems caused by the large-scale construction of roof distributed photovoltaic power stations in the future, and realize the group control of photovoltaic power stations, an intelligent Internet of Thing (IoT) communication system, which based on high-reliability power line carrier communication and 5G wireless communication technology, is proposed in this paper. Meanwhile, for solving the problem of resource collision and critical data loss caused by a large number of intelligent fusion terminals accessing the wireless network, a priority-based random access congestion control algorithm is proposed. By performing priority grouping, the algorithm allocates the backoff window dynamically. Through experimental simulation, it can be seen that compared with the classic binary backoff algorithm, this algorithm is able to significantly improve the stability of data transmission, reduce the packet loss rate, and play a role in alleviating network congestion and optimizing network performance.

The high-frequency link matrix converter (HFLMC) is composed of a traditional H-bridge circuit connected with the matrix converter through a single-phase high-frequency transformer (HFT). Compared with the traditional isolated converter, the high-frequency link matrix converter not only realizes the isolated DC-AC conversion, but also improves the transmission efficiency. However, the SVPWM modulation method for high-frequency link matrix converter will have relatively large circulating currents. In this paper, the reason of circulating currents of high-frequency link matrix converter is analyzed. Based on space vector modulation, an improved SVM modulation on method is proposed. A zero vector state is added between the two active vectors to reduce the circulating currents generated when switching between the active vectors. The working mode of the improved SVM modulation method and the active vector action time are analyzed. Finally, the correctness of the modulation method proposed in this paper is verified by simulation.

With the expansion of power grid scale, the performance bottleneck of traditional reliability assessment methods has gradually become prominent. How to improve the efficiency of reliability assessment has become an urgent problem to be solved. At the same time, the widespread access of distributed generation in the distribution network also affects the reliability of the distribution network. In view of these problems, this paper proposes an improved Warhall topology analysis method and an improved Monte-Carlo method to record fault sections, which improve the efficiency of reliability evaluation. On the basis of the improved method, the reliability index of a distribution network with DG access is calculated through a specific example, and the influence of DG access and access location on system reliability is further analyzed.

In order to cope with the increasingly complex situation of safe operation of power grid, this paper proposes fault event reasoning of substation power grid based on knowledge map technology. By using the method of knowledge map, the logical relationships such as co-reference relationship, causality relationship and time sequence relationship among substation monitoring events are established, and the rules and patterns among the events are described. Based on the power grid equipment entity and concept map, business logic map and historical case map, and according to the key information flow after fault signal analysis, the auxiliary decision of fault handling operation mode is made by using power grid operation and control logic, rules and experience knowledge. Realize the substation power grid fault analysis and processing function, and further improve the intelligent level of fault management.

The electromagnetic loop network problem of power system is a kind of power network problem that evolves with the development of power grid. In order to avoid the further spread of the problems caused by the electromagnetic ring network, the open-loop operation mode is widely used in the power grid below 500kV to reduce the impact on the power system. In electrified railways, the same phase connection of adjacent traction substations can be achieved through bilateral power supply technology, which increases the terminal voltage level of the traction network. However, an electromagnetic loop network is formed at the 27.5kV traction feeder. The purpose of this paper is to study the generation mechanism of electromagnetic ring network and its influence to power system under bilateral power supply technology. Theoretical support and technical guidance are provided for engineering practice and trial operation.

With the development of the field of scientific computing, the rapid generation of high-quality random numbers has become a crucial aspect of research. However, traditional methods are often insufficient in meeting the efficiency and quality requirements of random number generation. In this paper, we propose an improved version of the Mersenne Twister algorithm, implemented on the FPGA platform. By optimizing the characteristic polynomial in the Transform module, the randomness and uniformity of the Mersenne Twister (MT) algorithm are enhanced. The core modules of the algorithm are realized through a parallel structure one the FPGA, with the AXI4-Stream data bus utilized as the protocol for data interaction with peripherals. Finally, the quality of the random numbers by the designed generator is tested, and the results demonstrate that it can efficiently produce high-quality, uniformly distributed random numbers. These numbers are well-suited for applications that require the rapid generation of large quantities and high-quality random numbers, such as Monte Carlo simulations.

Wireless charger is a reasonable power supply option for electric vehicles (EVs), especially the multi-receiver one system showing potential in the charge-while-drive scenario. From the circuit analysis, it can be seen difficult to maintain specified power distribution among receivers. This study then proposes a real-time communication-free power flow management method of the multi-receiver wireless charger for EVs, considering constant power distribution among receivers. The control diagram and detailed control process are explained. Take the dual-receiver system as an example, constant power distribution is reached and the proposed method has been validated.

The miniaturization and lightweight of wireless EV chargers are of great significance. For the inductive power transfer system based on double-sided LCC compensation topology, an integrated design is proposed in which all the magnetically coupled components and compensation components are integrated into the aluminum shell of wireless chargers. For the quadrature arrangement of DD coils, including the main inductor and compensation inductor, the magnetic field distribution of the magnetic coupler is analyzed, the influence of ferrite thickness is studied, and the optimization method of ferrite shape is presented. Under the condition that the magnetic coupling effect is very little affected, the weight of ferrites in the proposed optimized scheme is reduced by 47% compared with the initial scheme. In the prototype experiment, the DC-DC efficiency reached 95.38% when the output power of the system was 6.93 kW.

In this paper, a compensation network parameter design method based on multi-objective optimization algorithm has been proposed to meet the robust requirements of the output characteristics of wireless power transfer (WPT) system under non-resonant conditions. The equivalent circuit mathematical model of WPT system under non-resonant condition is established, and the sensitivity of system output characteristics to different component parameters is analyzed. A compensatory topology parameter optimization method of double-sided LCC based on NSGAII algorithm is proposed, which effectively reduces the system output power fluctuation when multiple parameters fluctuate together, improves the output characteristics of the double-sided LCC WPT system, and enhances the robustness of the system output. Finally, the proposed method is validated by comparing the power variation before and after the system parameter optimization under the condition of multi-parameter combined fluctuation and the condition of sensitive parameter random fluctuation. It has lower power variation when component parameter fluctuates randomly and it improves the controllability of the system to a certain extent.

In order to ensure the stable operation of the power system, conventional voltage regulation methods include adjusting the generator excitation current and voltage regulation, adjusting the transformer turns ratio and voltage regulation, using SVC, STATCOM, synchronous phase shifter, etc. In terms of frequency regulation, the primary frequency regulation method of reducing the active power of multiple generators with lower frequency is adopted, or the secondary frequency regulation method using hydropower plants and medium-temperature and medium-pressure thermal power plants as frequency regulation plants is used. In terms of peak load regulation, thermal power units are often used as flexible peak regulation power sources, and the power output of thermal power units is controlled by load changes to achieve power balance. New power systems have the characteristics of large-scale integration of new energy, and random fluctuations in power on both the power generation and load sides. Energy storage, as a powerful new regulation means, is showing significant advantages with the support of national policies. This article first analyzes the energy storage technology-related policies issued by the government, and, combined with the characteristics of electrochemical energy storage technology in power auxiliary services such as voltage regulation, frequency regulation, and peak regulation in new power systems, puts forward prospects for the future application of electrochemical energy storage technology.

With the increase of the coverage area of the domestic distribution network, the probability of failure of the distribution network system is also greatly increased. In this paper, a grounding fault line selection system is designed for the small current grounding system of the distribution network. Firstly, taking the fault recorder as the data source, a single-phase grounding fault line selection system is built to monitor the single-phase grounding fault of the small current grounding system in the distribution network of each substation and select the fault line accurately and quickly. Finally, a 380 V physical simulation platform is built to simulate the fault line selection test. Taking Changchun area as an example, a single-phase ground fault occurs in the 66 kV system of Jingyang substation. Through the analysis of the system, the feasibility and accuracy of the line selection system are verified

In this paper, a phase selection method for classifying ten types of short-circuit faults in nuclear power plant is proposed. This method can quickly and accurately realize the short-circuit faults phase selection of nuclear power plant. Different from the conventional phase-mode transformation to the unmeasurable current fault components as the data source, the current fault components of measurable nodes herein as a data source. First of all, the amplitude and phase relationship between the current fault components at the measurable nodes and the unmeasurable current fault components at the fault points are compared by simulation. Secondly, the basis for phase selection is obtained by analyzing the boundary conditions of various short-circuit faults. Thirdly, the adverse effect of the noise of the measurable current fault component on the accuracy of fault selection is eliminated by Fourier transform, and the accuracy of fault phase selection is improved. In this paper, the MATLAB simulation verification is carried out through the IEEE13 nodes test feeder, which can well realize the short-circuit faults phase selection of the nuclear power plant. This method takes the current fault components at measurable nodes as the data source and can be applied in the actual nuclear power plant of power systems.

Lightning arrester in the power system to assume a very important role, its thermal characteristics are directly related to the arrester can operate normally. Surge arrester in addition to long-term frequency voltage also intermittently withstand the frequency overvoltage, lightning overvoltage and operating overvoltage and other transient overvoltage, resulting in the rise of the internal zinc oxide valve temperature. If the heat cannot be distributed in time, the valve temperature exceeds a certain limit value will cause thermal damage to the arrester [1, 2]. This paper investigates the thermal characteristics of the arrester by using the half-boundary method. By solving the differential equation for thermal conductivity in the cylindrical coordinate system, simulations were performed to study the thermal characteristics of the arrester under two-dimensions. The thermal characteristics of the arrester are obtained, and the research results provide an important reference for the stable operation of the arrester.

Aiming at the problem of pedestrian targets occlusion and multi-scale error and missed detection in pedestrian detection, a lightweight pedestrian detection algorithm based on improved EA-YOLOv5n is proposed. This method introduces the ECA attention module into the backbone feature extraction network, and learns the channels of pedestrian images by learning Information, improve the accuracy of pedestrian object detection in the case of occlusion, improve the calculation method of Bounding box loss function for the disadvantages of loss function calculation, adopt EIoU Loss and introduce power transformation to obtain higher bounding box regression accuracy. The experimental results show that using the improved model to conduct experiments on the Widerperson dataset reaches 69.6% mAP, which is 2.0% higher than the original algorithm, and the detection speed reaches 65FPS.

Wind turbine gearbox plays an important role in wind power generation equipment, and improving its reliability can help energy security and ecological environment improvement. This paper introduces the development of wind power generation and the reliability index system of wind turbine, analyzes the typical problems of wind turbine gearbox, and further puts forward the methods to improve its reliability, so as to help the maintenance personnel of wind turbines find the abnormal operation of gearbox in time and take corresponding measures to ensure the normal operation of wind turbine.

Wind power generation is in a period of rapid development due to its mature technology and environmental friendliness. Wind turbine is the main equipment for wind power generation, and the performance of its gearbox determines the stability and economy of power generation. This paper introduces the changes of installed capacity in various stages of wind power generation in the world, analyzes the common failure forms of wind turbine gearboxes, and on this basis, studies the failure parts and main causes, and further puts forward preventive measures for the failure forms of gearboxes, which can provide reference for wind farm personnel to carry out maintenance of wind turbines

This essay investigates the lateral dynamics control problem of an intelligent car with respect to bounded disturbances. A sliding mode controller is designed to address the above issue by introducing an event triggering mechanism. Compared with existing lateral motion adjustment algorithms with periodic control, unnecessary signal samplings, transmissions, computations and actuations are avoided, which indirectly saves the limited energy. In addition, finite-time convergence performance is achieved due to the sliding mode. Finally, numerical simulation experiments are used to illustrate the theoretic results

In the past ten years, heart disease has been the main cause of death among Chinese residents. At present, the more accurate way to diagnose heart disease is invasive examination - cardiac angiography. This diagnostic method may cause serious arrhythmia, and some people may be allergic to contrast agents. Therefore, certain manpower and material resources are required to monitor the patient's vital signs after angiography. So, if we can use other patient data information to predict whether a person has heart disease through machine learning, it will make a great contribution to the prevention and diagnosis of heart disease. For this reason, this paper proposes an SVM-GBDT hybrid model based on feature selection to predict the occurrence of heart disease. After data processing, the regression results are obtained from the SVM model, and then the important attributes are filtered through feature selection by setting variance thresholds. The regression results are combined with the results of feature selection, and the GBDT model is used for prediction analysis. The experimental results show that the svm-gbdt hybrid model presented in this paper performs better than the single model at multiple evaluation metrics. When compared with the prediction effect of other machine learning models, the hybrid model proposed in this paper also performs well. As a result, the SVM-GBDT hybrid model based on feature selection proposed in this paper can play a helpful role in the prediction and diagnosis of heart disease.